7European Southern Observatory, Karl-Schwarzschild-Straße 2, D-85748 Garching bei München, Germany
8Max-Planck-Institut für Radioastronomie, Auf dem Hügel 69, D-53121 Bonn, Germany
9Astronomy Department, King Abdulaziz University, P.O. Box 80203, Jeddah 21589, Saudi Arabia
10School of Mathematical and Physical Sciences, University of Newcastle, University Drive, Callaghan NSW 2308, Australia
11Astrophysics Research Institute, Liverpool John Moores University, 146 Brownlow Hill, Liverpool L3 5RF, UK
12European Southern Observatory, Alonso de Córdova 3107, Vitacura Casilla 763 0355, Santiago, Chile
13Joint ALMA Observatory, Alonso de Córdova 3107, Vitacura Casilla 763 0355, Santiago, Chile
14Department of Physics and Astronomy, UCLA, 430 Portola Plaza, Los Angeles, CA 90095-1547, USA
15Department of Physics and Astronomy and CIERA, Northwestern University, Evanston, IL 60208, USA
16Department of Physics, CCIS 4-181, University of Alberta, Edmonton, Alberta, T6G 2E1, Canada
17Department of Physics and Astronomy, Kagoshima University, 1-21-35, Korimoto, Kagoshima, 890-0065, Japan
18International Centre for Radio Astronomy Research, Curtin University, Bentley, 6102, Australia
19Harvard-Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge, MA 02138, USA Received 2017 July 5; revised 2017 October 17; accepted 2017 October 18; published 2017 November 20
Abstract
The Survey of Water and Ammonia in the Galactic Center(SWAG) covers the Central Molecular Zone (CMZ) of the Milky Way at frequencies between 21.2 and 25.4 GHz obtained at the Australia Telescope Compact Array at
∼0.9 pc spatial and ∼2.0 km s−1 spectral resolution. In this paper, we present data on the inner ∼250 pc (1°.4) between SgrC and SgrB2. We focus on the hyperfine structure of the metastable ammonia inversion lines (J, K )=(1, 1)–(6, 6) to derive column density, kinematics, opacity, and kinetic gas temperature. In the CMZ molecular clouds, we find typical line widths of 8–16 km s−1 and extended regions of optically thick (τ>1) emission. Two components in kinetic temperature are detected at 25–50 K and 60–100 K, both being significantly hotter than the dust temperatures throughout the CMZ. We discuss the physical state of the CMZ gas as traced by ammonia in the context of the orbital model by Kruijssen et al. that interprets the observed distribution as a stream of molecular clouds following an open eccentric orbit. This allows us to statistically investigate the time dependencies of gas temperature, column density, and line width. We find heating rates between ∼50 and
∼100 K Myr−1 along the stream orbit. No strong signs of time dependence are found for column density or line width. These quantities are likely dominated by cloud-to-cloud variations. Our results qualitatively match the predictions of the current model of tidal triggering of cloud collapse, orbital kinematics, and the observation of an evolutionary sequence of increasing star formation activity with orbital phase.
Key words: Galaxy: center– evolution – ISM: clouds – ISM: kinematics and dynamics – stars: formation
1. Introduction
The Galactic Center (GC) and the Central Molecular Zone (CMZ) in particular represent an environment with conditions that are not to be found anywhere else on large scales in the Milky Way. The CMZ received its name from the presence of a large reservoir of dense (104cm−2) and molecular gas of a few times 107Me (Oka et al.1998; Morris & Serabyn1996;
Ferrière et al.2007) and covers the central ∼500 pc of the GC region. The large amount of molecular gas is found to be accompanied by a relatively high star formation rate (SFR) of
∼0.1 Meyr−1(Longmore et al.2013a; Barnes et al.2017). Star formation(SF) laws that build on the assumption of a constant gas-depletion time of 1 Gyr can fit the observed SFR (Bigiel et al.2010; Leroy et al.2008,2015). However, these relations are derived from gas at much lower densities than observed in
the CMZ. Density-dependent SF laws, on the other hand, strongly overpredict the SFR (Longmore et al. 2013a) at
∼0.4 Meyr−1 (Kennicutt 1998; Krumholz & McKee 2005;
Krumholz et al. 2012) to ∼0.8 Meyr−1 Lada et al.
(2010,2012). Thus, the star formation efficiency (SFE) in the CMZ is significantly lower than expected for the observed gas densities. The complex interplay of energetic processes in the GC allows for different potential answers to this problem of low SFE, but none of these processes alone can explain the discrepancy (Kruijssen et al. 2014). A general picture of episodic starbursts in gas rings in galactic centers introduced by Krumholz & Kruijssen (2015) and Krumholz et al. (2017) might solve the SFE problem in the particular case of the CMZ and set the framework for another intriguing feature of the GC:
a ring-like structure of dust and molecular gas. This structure
(see Figure1for an overview) is projected onto an infinity (¥) shape that follows several arcs, the most prominent being the so-called “dust ridge” stretching from the massive molecular cloud G0.253+0.016 (“the Brick”) to the star-forming region Sgr B2(Lis et al. 1999). It might be continued via an[l+,b-] arc and another loop at negative longitude ([l-,b+],[l-,b-]) passing through the star-forming region Sgr C. The syntax
+ - + -
[l ,b ]denotes the four quadrants in Galactic coordinates at positive/negative Galactic longitude (l) and latitude (b).
Clouds at[l-,b-]and[l+,b+]are thought to be located in front of SgrA*, i.e., the near side of the GC that is visible in silhouette against the background, whereas most of the[l-,b+] and[l+,b-]gas is more likely to be on the far side, i.e., behind SgrA*(Bally et al.2010; Molinari et al.2011).
Various models of the gas and dust distribution in the GC exist: a simple bar model(Morris & Serabyn1996); variations of a spiral arm model in which the apparent ring is formed by the inner part of two spiral arms (Sofue 1995; Sawada et al.2004; Rodríguez-Fernández & Combes2008; Rodríguez- Fernández2011; Ridley et al.2017); a closed, twisted elliptical ring detected in dust emission (Molinari et al. 2011), and a sequence of open-ended gas streams (Kruijssen et al. 2015, hereafter K15).K15highlight the impossibility of closed orbits in extended gravitational potentials and provide a better fit to single-dish ammonia emission in position–position–velocity (PPV) space, as was confirmed by Henshaw et al. (2016b) for three molecular species in single-dish observations. Thus, we focus on this model and call it the“stream model” in contrast to the“ring model” and the “spiral arm” model. According to the stream model, the stream of GC molecular clouds oscillates radially and vertically, and can be traced for∼1.5 revolutions around the GC. The radial oscillation periodically brings dense molecular gas closer (r∼60 pc at pericenter) to the gravita- tional center (traced by Sgr A*) and deeper into the potential, whereas the apocenter lies at r∼120 pc. As the CMZ clouds slowly evolve toward low virial ratios as the turbulent energy dissipates (Krumholz & Kruijssen 2015; Walker et al. 2015;
Henshaw et al.2016a), it is statistically more likely that cloud collapse occurs at the pericenter where the compressive tidal forces are strongest along the orbit and thus a cloud receives thefinal nudge for the transition to self-gravitation. Subsequent
SF stages will then occur downstream from the pericenter passages and could potentially be observed as an SF sequence if the orbit is sufficiently sampled with molecular clouds that start to collapse at a similar point of their orbits. This model of triggered star formation wasfirst proposed by Longmore et al.
(2013b) based on the observation of different evolutionary SF stages along the dust ridge. The two young stellar clusters in the GC, Arches and Quintuplet, may also support this model as their orbits and ages are consistent with formation at a common point after collapse of their respective parent clouds at a pericenter passage (Figure 1 of Stolte et al. 2014; Kruijssen et al. 2015). In addition to stars and star formation tracers, cloud properties are also expected to show evolutionary behavior along the gas streams; this has not been thoroughly tested yet. Weak hints toward rising gas temperatures in the dust ridge are suggested by Ginsburg et al. (2016), while a recent paper by Kauffmann et al.(2017b) based on N2H+data from the Galactic Center Molecular Cloud Survey can neither confirm nor exclude the possibility of triggered evolution when examining mass-size relation and SF suppression. Both analyses were based on a low number of measurements(∼35 measurements in dust ridge clouds in Ginsburg et al.2016and six clouds in Kauffmann et al. 2017b) and lack the statistical power to detect potential evolution in the presence of scatter.
In this paper, we aim to test the prediction of the stream model by searching for coherent temporal evolution of molecular gas properties on a broad statistical basis using the Survey of Water and Ammonia in the Galactic Center(SWAG). SWAG targets 42 molecular and atomic species at∼21–25 GHz, of which at least 20 are detected, and thus it offers an ideal database for such an analysis. SWAG maps the CMZ in the region of∼−1° to
∼+2° in Galactic longitude at latitudes∣ ∣b < 0 . 4 using the Australia Telescope Compact Array (ATCA) and achieves a spectral resolution of 0.4 km s−1at∼20″ spatial resolution. One of the survey’s signature targets is the ammonia molecule, the properties of which are well suited to exploring the thermal and kinematic structure of the interstellar medium (ISM; Ho &
Townes 1983). In an ammonia (NH3) molecule, the nitrogen atom can tunnel through the plane of hydrogen atoms(inversion) at a rotational-state-dependent frequency that can be related to temperature. These inversion frequencies are closely spaced
Figure 1.Overview of the GC. The gas(background, SWAG NH3(3, 3) peak intensity) resembles the shape of an infinity symbol, ¥, and is parametrized as a sequence of stream segments byK15. The direction of motion is along the dust ridge([l+,b+]denoting positive longitude and latitude) to Sgr B2,[l+,b-],[l-,b+]to Sgr C and[l-,b-]. Gas in the dust ridge[l+,b+]and[l-,b-]is mostly on the near side(in front of the GC), whereas[l+,b-]and[l-,b+]gas is mostly on the far side (Bally et al.2010; Molinari et al.2011). The most important sources are labeled.
emission 1 4 (3.3 pc at 8.3 kpc)20 is filtered out by the ATCA interferometer in this setup. Spectral fitting of the ammonia hyperfine structure allows us to construct maps of line- of-sight velocity, line width, opacity, and kinetic gas temperature that can be tested for evolutionary behavior under the assumption of a kinematic model. Applying this method to SWAG data and the stream model allows us to check gas properties for absolute time dependencies that can be expected if a sequential star formation sequence is present in one or more of the stream segments.
As the structures in the GC are complex, simple names can be misleading. Throughout this paper, we refer to the“ring” as the general ¥-like structure of CMZ gas without being interested in the detailed structure or kinematics. “Stream,”
however, refers to theK15model that introduced a stream as a set of four segments of an orbit wrapping around the GC whose subdivision does not have physical meaning but simplifies descriptions.
In this paper, we summarize the observational setup and data reduction of the SWAG survey in Section 2 on the basis of observations of one-third of the total area, covering the region between Sgr B2 and Sgr C. Section3shows and discusses the primary data products (channel and moment maps, spectra, position–velocity diagrams), and Section4describes thefitting of the ammonia hyperfine structure inversion lines. In Section5 we use the derived thermal and kinetic properties of GC molecular clouds to discuss their properties in the context of the orbital model of K15 by reconstructing potential time dependencies of these properties. The resulting relations are discussed in the context of the star formation sequence as proposed by Longmore et al.(2013b). A summary in Section6 concludes this work. Appendices show further SWAG data products (AppendixA), details on imaging (AppendixB), the ammonia thermometer (AppendixC), and hyperfine structure line fitting (Appendix D), as well as further fit-derived data products(AppendicesEandF) and further time evolution plots (AppendixG).
2. Observations and Data Reduction
SWAG is a large survey aiming to map the entire CMZ of the Milky Way from ∼−1° to ∼+2° in Galactic longitude at latitudes∣ ∣b < 0 . 4. The boundaries correspond to an integrated surface brightness level of 0.1 K km s−1 of NH3 (3, 3) from
2.1. Observations
The observations for SWAG were carried out with the ATCA21interferometer. Achieving the maximum sensitivity to extended emission requires the most compact array configuration, H75, with baselines of 31–89 m between five dishes. A sixth antenna that is not considered in this work provides baselines of
∼4.4 km to the inner antennas. The primary beam FWHM of the 22 m dishes at ν=23 GHz is ∼2 4. Every position in the targeted area is covered by the primary beam of at least three pointings. In the radio K-band regime, the H75 array configura- tion is not sensitive to emission extending over more than∼1 4 (3.3 pc) because of interferometric spatial filtering.
Between 40 and 220 pointings are combined into strips of
∼4 5 width, as shown in Figure2, which are observed between subsequent observations of the phase calibrator. Long strips are split into Galactic northern and southern parts to ensure frequent phase calibrations. In order to maximize the u v, coverage, pointings are not observed one after the other, but in rows of even and odd numbers, as explained in Ott et al.
(2014). The rows are aligned parallel to Galactic longitude (“l-scan” in Ott et al.2014), while the first element of each row determines its parity (odd, even). Rows of odd parity are observed consecutively, followed by the left-out rows of even parity at somewhat later local sidereal times (LST) and therefore other projected baselines, which increases the u v, coverage. The resulting typical u v, coverage of a single pointing is shown in Figure3.
The Compact Array Broadband Backend (CABB, Wilson et al.2011) provides two IF bands, each of 2 GHz bandwidth.
In the CFB 64M-32k mode adopted for these observations, each 2 GHz band was divided into 32×64 MHz “continuum”
channels with 16 “zoombands” of 2048×32 kHz channels selected in each IF band for spectral line observations. The integration time of each pointing was set to 8×30 s=
4 minutes as a trade-off between low noise and total project time, but it varied due to reobservations andflagging. The eight individual integrations were scheduled for best u v, coverage to be observed over the whole available range of LST from about 15:00 to 22:00 hr as much as possible. Owing to pointing overlap, the actual sensitivity is higher by a factor of 3 ( 2 for edge pointings) than can be expected from the integration time of a single pointing. The expected noise calculated from the radiometer equation s µ
nt D Tsys
as a function of system temperature Tsys, bandwidth Δ ν, and integration time τ is
20We adopt this distance measurement by Reid et al.(2014) throughout the paper. The IAU recommends to use 8.5 kpc, while a recent study by Boehle et al.(2016) found 7.86 kpc.
21The ATCA is part of the Australia Telescope National Facility, which is funded by the Australian Government for operation as a National Facility managed by CSIRO.
σ=28 mK for a typical Tsys=80K in one channel of Δν=32 kHz (∼0.4 km s−1 at 23 GHz) and t=240 s. This value corresponds to σ∼6.0 mJy beam−1 for NH3 (3, 3) (ν=23.87 GHz, beam 26 03×17 71), which is about half of the measured noise of ∼13.1 mJy beam−1 (Section 2.3.1) because the deconvolution is imprefect owing to the sparsely sampled u v, plane.
Observations were scheduled in three-week blocks in the southern hemisphere winter months of July/August in each year from 2014 to 2016. The total project time including calibration data sums up to ∼720 hr (∼204 hr for the data presented in this paper), including additional time needed to reobserve some regions whose observations were affected by
bad weather, resulting in increased noise. The total on-source time was∼525 hr (∼150 hr shown in this paper).
Bandpasses and delay calibrations for each antenna are derived from daily 10-minute integrations on PKS 1253-055 (3C 279). Its flux of ∼16–20 Jy at frequencies around 23 GHz is bright enough to derive accurate bandpass solutions. The gain/phase (complex gain) calibrator PKS 1710-269 was observed for 2 minutes every 20–30 minutes, depending on thefield, after each l-scan and in between l-scans if necessary.
Daily flux calibrations of 5-minute integration time are performed on the radio galaxy PKS 1934638, which is the standard primary flux density calibrator for the ATCA below 50 GHz.
2.2. Calibration
All steps from data import to imaging are performed with the ATCA specific version of the package MIRIAD22 (Sault et al. 1995). The necessary steps were developed into a data reduction pipeline that provides the base for reducing the entire survey.
2.2.1. Flagging
Raw data are imported via atlod with options birdie, nocacal, noif, and opcorr to automaticallyflag resonant instrument modes(birdies), initial array calibration data (cacal), set correct intermediate-frequency mapping behavior, and to apply atmospheric opacity corrections. Some of the array setup data were not detected by the automatic routine and had to be flagged manually with uvflag.
Resonant modes in ATCA’s backend system cause addi- tional birdies in channelsn ´1024+1(n=1, 2,¼)that are not identified during data import. The missing channels are interpolated later when averaging and Doppler-correcting the visibilities to a common resolution of 2 km s−1. To remove additional correlator artifacts and extreme RFI, calibrator data were clipped at 200 Jy and CMZ observations at 10 Jy.
Additional flagging of bad data was made by eye in the interactive task blflag. Polarization measurements were not required for the presented analysis, and all crosshands/cross- polarizations (XY and YX for ATCA’s linear feeds) were flagged. Antenna ca06 offers high spatial resolution at base- lines of >4300 m, but was also flagged due to lack of
Figure 2.SWAG observing layout. The spatial coverage of SWAG is derived from the 0.1 K km s−1contour(blue) of the Mopra single-dish CMZ survey by Ott et al.
(2014; background gray-scale image). Between 40 to 220 pointings are included in an observation strip (numbered from west to east and alternately colored). This work includes data of the inner CMZ covered by strips 10 to 30.
Figure 3.Typical u v, coverage of a single pointing of SWAG using the inner array of five antennas (excluding the far out antenna ca06). Note that overlapping pointings increase the u v, coverage by a factor of 3 ( 2 for edge pointings), which is not included in this plot. The u v, coverage was optimized by the scanning technique (“l-scan”) shown in Figure 1 of Ott et al.(2014).
22http://www.atnf.csiro.au/computing/software/miriad
intermediate u v, points, which drastically down-weights these long baselines. Except for the instrumental birdies we described, little radio frequency interference (RFI) is present at frequencies of∼23 GHz.
2.2.2. Complex Gain Calibration
Bandpass amplitudes and phases are calculated on the designated bandpass calibrator PKS 1253-055. The resulting correction is then applied to the gain/phase calibrator PKS 1710-269 and flux density calibrator PKS 1934-638, and complex antenna gains are derived with mfcal.
Calibration solutions were inspected visually to identify remaining problems that were subsequentlyflagged. Recalibra- tions were performed until well-behaved solutions were found.
2.2.3. Flux Density Calibration
The calibration quality is assessed with Figure4. The phase calibrator PKS 1710-269 was imaged including all calibration corrections for all observation days and all zoombands and the peakflux density listed. Measured flux densities per zoomband are constant within ∼15%, which is typical for radio observations at this frequency. The origin of this variability cannot be identified without absolute measurements and can be due to changing atmospheric conditions or intrinsic luminosity variation in the QSO PKS 1710-269. Thus, 15% is an upper limit on the flux density variation across observation days, which is similar to the absoluteflux uncertainty.
Spectral lines observed in the same zoomband, however, share a common flux density uncertainty, and thus flux ratios and derived quantities are more accurate. This provides a robust basis for the derivation of the kinetic gas temperature, which is primarily determined from ammonia line ratios.
2.2.4. Continuum Subtraction
Over the relatively small spectral range of a zoomband of
∼64 MHz, a linear approximation captures the spectral index of any continuum emission well enough. Line-free channels could not be identified in visibility spectra because the brightness of most of the spectral lines is too low; instead, they are calculated from the line’s rest frequency and a window of ±400 channels (corresponding to ∼178–150 km s−1at 21.0–25.0 GHz), which is large enough to exclude the typical velocity range of molecular gas in the CMZ of ±150 km s−1. Any possible emission outside these line-free windows is weak and does not affect the continuumfit. Fit and subtraction were made using the uvlin task, forcing afirst-order polynomial.
Rest frequencies are set according to Table 1, which lists values from splatalogue.23
Merging five channels to obtain a spectral resolution of 2 km s−1increases the signal-to-noise ratio(S/N) by a factor of 5 ~2.2, while the spectral resolution is still adequate for the expected typical line widths of 5–25 km s−1 in ammonia emission.
2.3. Imaging 2.3.1. Deconvolution
Fourier transformation from the visibility to the image domain is performed with the task invert, which includes primary beam correction and combines the visibilities of single pointings into a mosaic. The Briggs weighting parameter is set to +2, which is close to natural weighting. The cell size(pixel size) is set to 5″ in R.A. and decl., which oversamples the minor axis of the synthesized beam (24 8×16 9, 89°.3 at 25.1 GHz) by a factor >3. At lower frequency, the beam size grows up to 26 2×17 8 at 23.7 GHz, which results in a
Figure 4.Flux density of the phase calibrator PKS 1710-269 per observation day after all calibration steps for the 30 frequency bands of CABB. On observation days 24 and 29, correlator crashes required new calibrations, which are labeled a and b and correspond toflux densities before and after the crash, respectively. Flux densities are constant across observation days within∼15%. The increasing values toward lower frequency is an intrinsic property of the source: a quasar with spectral index ~-0.5. The precision of theMIRIADtask imstat is two decimal places, which leads to quantized data.
23http://www.splatalogue.net
variation of linear size of∼5% between NH3(1, 1) and (6, 6).
These beam size are∼1.7 and ∼1.2 times the naively expected size of∼15″×15″ due to weighting. Table1lists the obtained beam sizes for NH3(1, 1) to (6, 6).
Each pointing is imaged over 512 pixel(FOV 42 7) and then integrated linearly into the mosaic with another weighting function that smooths noise across regions of differing pointing coverage (MIRIADdefault method, Sault et al.1996).
The extended emission of most spectral lines is generally better deconvolved by an algorithm that also uses extended sources for modeling. The only possibility of extended modeling that MIRIADoffers is the maximum entropy method (MEM). Our tests with the mosaic versions of MEM (mosmem) and clean (mossdi) on SWAG data confirmed that more extended emission was reconstructed by MEM over regular clean.
A first run with up to 10 iterations deconvolves the dirty image well enough to construct a mask that prevents treating regions without significant emission. Image restoration is done with restor. A mask is calculated to contain all pixels with S/N above ∼5, which corresponds to 65.0 mJy beam−1in the semi-cleaned image.
Deconvolution is then repeated with 50 iterations in mosmem only inside the pixels that were set as relevant by the mask. The restored images still contain sidelobe structures, especially around strong sources and in strong lines. The difficulty of obtaining better images with mosmem is described in more detail in AppendixB.
The root mean square(rms) noise is not constant within and across the data cubes, but increases slightly with decreasing frequency because of the proximity to an atmospheric water line that increases the telescope’s system temperature. At a level of ∼13.0 mJy beam−1, the difference between the channels at −150 km s−1 and +150 km s−1 relative to rest frequency is 0.1–0.6 mJy beam−1 for the ammonia lines listed in Table1. Thus, a common value is calculated as the mean of minimum and maximum noise measurements. The spatial variation of noise across pointings is negligible except for the two outermost strips of the data considered in this work(10/11 and 30/31, see Figure2), which show increased noise levels in
∼50″ and
∼100″ (∼one-third and two-thirds of the primary beam, respectively) as the smaller radius of 50″ proved to be enough to exclude any visible higher noise edges in ammonia distribution maps, whereas spectral fitting is affected by increased noise to∼90″ from the edge.
2.3.3. Moment Maps
Data cubes were collapsed along the spectral axes with the task moment in MIRIAD to obtain moment maps. Map edges are blanked with the 50″ edge mask and noise is excluded at levels of 3σ and 5σ for intensity maps (peak intensity; moment 0, integrated intensity) and velocity maps (moment 1, intensity-weighted mean velocity; moment 2, velocity dispersion), respectively.
2.3.4. Position–Velocity Diagrams
Position–velocity cuts through the restored datacube edge masked at 100″ were calculated with the task impv in the CASA software package24 (McMullin et al. 2007) along Galactic longitude, averaging over all emission in Galactic latitude.
3. Results
Figures in this section show NH3(3, 3) as representative of all six observed metastable ammonia inversion lines as it is typically the brightest ammonia line observed with SWAG.
The correspondingfigures for NH3(1, 1), (2, 2), (4, 4), (5, 5), and(6, 6) can be found in AppendixA.
3.1. Data Cubes
Channel maps of 2 km s−1 width in steps of 15 channels (30 km s−1) are shown in Figure5. All velocities are specified in the LSRK frame using the radio approximation. Emission can be detected in the range of ∼−180 km s−1 to ∼150 km s−1 with emission related to the ring-like feature between∼−150 km s−1 and∼120 km s−1. Extended gas clumps at l∼0°, b∼0°.15 are offset in velocity from the surrounding gas and are detected among the highest velocity gas in thefield of view. Typical flux density values for molecular clouds in theNH3(3, 3) line are on the order of a few hundred mJy beam−1 with peaks of 4.23 Jy beam−1 (19.6 K) in the 20 km s−1 cloud and 5.28 Jy beam−1 (24.8 K) in Sgr B2. The spatial distribution is very similar among the ammonia lines, despite the varyingflux density levels(Sections5andA.1).
NH3(6, 6) 25.05603 409.2 24.8×16.9 89°.3 13.2 16.2
mean value 13.0 17.6
Note. snormalis relevant for most of the maps beside the two outermost strips (see Figure 2) on each side that are described by shigh. Given values are an average over 20 (snormal) and 12 (shigh) measurements in two channels at
±150 km s−1relative to the rest frequency.
24Common Astronomy Software Applications,https://casa.nrao.edu/.
3.2. Sample Spectra
Depending on the structure of the source and the spectral line, hyperfine satellite components and multiple components along the line of sight are detected. A bright single-component example spectrum of NH3 (3, 3) is shown in Figure 6 toward the Brick
(G0.253+0.016). Absorption against continuum supposedly from free–free emission is detected toward the 50 km s−1 cloud, the Brick, clouds c and f, and some smaller clouds in thel+,b-and
- +
l ,b portions of the“ring” (cf. Figure1).
Further spectra at the same position in the Brick for NH3(1, 1) to(6, 6) are shown in AppendixA.2.
Figure 5.Channel maps of NH3(3, 3). Every fifteenth channel of 2 km s−1(separated by 30 km s−1) in the range of −120 km s−1to+120 km s−1is shown. Velocity and beam(26 03×17 71, 89°.3) are indicated in the top and bottom right corners, respectively. The conversion factor from flux density to brightness temperature for this beam size at 23.87 GHz isT [ ]K S[Jy beam-1]=4.65.
3.3. Image Moments
Image moment maps are shown in Figure 7 for orders 0 (integrated intensity), 1 (velocity centroid), and 2 (velocity dispersion); the top panel shows the peak intensity.
Peak intensity (Figure 7(a)) and integrated intensity (Figure7(b)) show an almost identical distribution of emission, with SgrC being a notable exception of strong peak but low integrated intensity, indicating narrow line widths. SgrB2, the arching structure of the dust ridge (0°.2<l<0°.6) and the sequence of massive clouds at −0°.2<l<0°.15 with a bifurcation25 at l∼0°.1 are most prominent in both intensity maps. The line-of-sight velocity(Figure7(c)) shows a complex pattern at 0°.0<l<0°.2 of multiple partially overlapping emission components along the line of sight. At negative longitudes, the coherent structure of negative velocities is intermingled with clouds at positive velocities. In total, the molecular arcs trace an apparently ring-like structure. As the interferometerfilters out diffuse emission, the molecular clouds in this ring-like structure must be rather clumpy, which has been reported previously (Bally et al. 2010; Molinari et al.2011; Ott et al.2014; Ginsburg et al.2016).
Using the definitionl+ -andb+ -denoting positive/negative values of Galactic longitude and latitude, the general sense of rotation in the CMZ is consistent with the scenario [l , 0- ]
- -
([l ,b ]→[0, 0]→[l+,b+]→[l , 0+ ]→[l+,b-]→[0, 0]→
- +
[l ,b ] found in previous studies (e.g., Molinari et al. 2011;
Kruijssen et al.2015; Ginsburg et al.2016). This, however, does not necessarily mean that arcs are physically connected, e.g., note the 50 km s−1jump between[l-,b-]and[l+,b+], but indicates that the bulk of the gas is following the same sense of rotation.
Typical velocity dispersions in the moment-2 map are found to be 5–25 km s−1, but velocity dispersions for individual clouds are expected to be lower because of multiple components along the line of sight, and for the lower excitation states, blending of hyperfine components.
Moment maps of NH3 (1, 1) to (6,6) are shown in AppendixA.3.
(inversion) depends on the rotational state(J, K ) of the molecule. The population of these states is temperature dependent according to a Boltzmann law. For each rotational state J (angular momentum), there exists another ladder of states with quantum number K(angular momentum projected onto the symmetry axis). Each state decays quickly down the K-ladder to the lowest energy state (J, K=J), e.g., NH3 (1, 1). These states are metastable as they can only deexcite through forbidden transitions if the density is too low for collisional deexcitation and thus live long enough to be commonly observable in the ISM. The inversion lines are further split into hyperfine structure lines by the interaction of the electric quadrupole moment of the nitrogen atom with the electric field of the molecule’s electrons, which allows the derivation of gas opacity from the ratio of hyperfine components. Fits of the entire line structure of at least two metastable ammonia lines thus yield opacity-corrected column density, line-of-sight velocity, line width, and rotational temperature. The rotational temperature is a lower limit and can be used to estimate the kinetic gas temperature up to approximately the line excitation energy via model-derived conversion functions.
Details on the ammonia thermometer, the derivation of opacity-corrected column density, rotational temperature, and its conversion into kinetic gas temperature are described in AppendixC.
4.1. Hyperfine Structure Fitting inCLASS
In order to derive gas temperatures without having to assume optically thin emission, we need to know line temperatures, line widths, and opacities. We derive these quantities byfitting the ammonia hyperfine structure with the CLASS26 package in GILDAS.27The 100″ edge-masked data cubes were masked at 3σ per channel and 5 Jy beam−1 km s−1 in integrated flux density to ensure sufficient S/N for successful fits. Fitting is made with the CLASS functions method nh3(j,j) and minimize for J=1, 2, 3. The higher transitions J=4, 5, 6 are not implemented but can befitted with the method hfs using the relative positions and strengths of the ammonia hyperfine structure components of Townes & Schawlow (1975;
listed in Table 3). The fitting algorithm is constrained to line opacities in the range 0.1τ30, which introduces a factor of
t -
- =
t
t
-
- ( )
e lim 0.1 1 e
1 1.051 1
0 0.1
Figure 6. Sample spectrum toward the Brick (l=0°.2328, b=0°.0107) extracted from the NH3(3, 3) cube. rms noise is 13.1 mJy beam−1(0.061 K).
The positions of the hyperfine satellites according to Table3are indicated by dashed vertical lines. Note the broad line width that blends the outer hyperfine components(which have separations of ±21.1 km s−1and±29.1 km s−1from the main line) into a pedestal.
25This term is adopted fromK15and denotes the apparent splitting after the 50 km s−1 cloud when moving toward positive longitudes, best seen in Figure5at 60 km s−1.
26http://iram.fr/IRAMFR/GILDAS/doc/html/class-html/
27https://www.iram.fr/IRAMFR/GILDAS/
Figure 7. NH3(3, 3) moment maps: (a) peak intensity, (b) integrated intensity (moment 0), (c) intensity-weighted mean velocity (moment 1), and (d) intensity- weighted velocity dispersion(moment 2). The intensity maps (a) and (b) are masked at 3σ, and velocity maps (c) and (d) are masked at 5σ with an rms noise of σ=13.1 mJy. The beam of 26 0×17 7 (1.05 pc×0.71 pc) is indicated in the lower right-hand corner of each panel.
relative to the line temperature of optically thin emission(τ → 0).
This deviation by up to 5.1% is small compared to the generalflux density uncertainties of ∼15% (Figure4) and can be neglected.
Thefit is further constrained to the strongest emission component along the line of sight. Occasionally, unsuccessful fits occur through blending of closely spaced emission components.
These are discarded based on reduced χ2 (CLASS parameter lineRMS > 0.1), physically implausible values (Dv FWHM( )>
50.0 km s−1), and the very large error of the fitted parameter (D(vlos)>10.0km s−1,Δ (Δ v)>10.0 km s−1). Examples and statistics of the hyperfine structure fitting can be found in AppendixD.
4.1.1. Sample Boltzmann Plot
A Boltzmann plot (Appendix C.1, cf. Goldsmith &
Langer 1999), typical for the ammonia emission in the GC, is shown in Figure 9. The observed shapes do not follow a single linear relation, but become more shallow around NH3
(3, 3) to NH3(4, 4), which can be due to the nonlinear Tkin–Trot
conversion or be indicative of a multiple temperature medium, as is already known in the literature (e.g., Hüttemeister et al. 1993; Mills & Morris 2013). Errors in derived column density are typically small (ΔNu/Nu on the order of a few percent) but larger (a few tens of percent) at the edges of clouds because of the lower S/N, which allows the derivation of accurate temperatures of typically 5%–10% relative error (Trot) and 10%–25% after conversion into kinetic temperature.
Additionally, the ∼15% flux density error (Figure 4) contributes to the column density errors.
4.2. Results of the Hyperfine Structure Fitting As in Section 3, only NH3 (3, 3) is shown to illustrate features common in all observed ammonia lines. As a representative temperature measure, we choose to show T24
calculated from NH3(2, 2) and NH3(4, 4). The strongest line, NH3 (3, 3), yields the temperature T36, which suffers larger relative errors than T24because of the low intensity of the NH3
(6, 6) emission. We also calculate T12 and T45. Other combinations, like T23, cannot be calculated because NH3
(2, 2) is a para state (one hydrogen nuclear spin antiparallel), while NH3 (3, 3) is an ortho state (all hydrogen nuclear spins parallel). Generally, NH3(i, i) with =i 3n(n integer) are ortho,
the others are para states. As the relative abundance of ortho versus para ammonia is not known, temperatures can only be estimated from two para states or two ortho states. For T12, the
-
Trot Tkinconversion(Table 4 and Figure 5 in Ott et al.2011;
see also Morris et al.1973) flattens, which prevents a reliable derivation of temperatures above∼50–60 K. T24, T45, and T36do not suffer such problems (Gorski et al. 2017) and yield very similar results. The corresponding maps for T12, T45, and T36can be found in AppendixE.
4.2.1. Column Density
The NH3(3, 3) column density Nl(Figure33) traces much of the ring-like gas stream by construction of the mask applied before hyperfine structure fitting. A peak of N33=7.7´1015
cm−2 is reached in SgrB2. Massive clouds like the Brick, G+0.10–0.08, and the 20 km s−1cloud are also found to have high column densities (N331.5´1015cm−2), while NH3
(3, 3) is not largely excited and detected at column densities
1.5´1015cm−2. The NH3 (1, 1) column density reaches a maximum of1.2´1016cm−2in SgrB2, where NH3 (6, 6) is still detected at up to2.6´1015cm−2.
The total ammonia column density as derived from Equation (22) (Figure 10) shows a similar picture as the NH3
(3, 3) column density. The highest column densities are reached in SgrB2 (6.8´1016cm−2), the Brick (3.9´1016cm−2), and the 20 km s−1 cloud (3.0´1016cm−2), whereas most smaller clouds along the “ring” have Ntot1.0´1016cm−2. The clouds at l~ 0 . 1,b~ 0 . 2 show the lowest detected column densities(1.0´1015Ntot cm-22.5´1015) of clouds that do not fall below thefitting thresholds. Other clouds of this size typically exhibit column densities that are higher by a factor of 2–3. A small cloud of high column density is located at l=−0°.40, b=−0°.22 with Ntot up to 2.6´1016cm−2 (N33=6.4´1015cm−2) and surrounded by gas of Ntot~ 1016cm−2. This cloud is assumed to lie in the foreground (Longmore et al.2013b) and is discussed to be influenced by an
Figure 8.Position–velocity diagram along Galactic longitude, integrated over Galactic latitude.
Figure 9.Typical Boltzmann plot(excitation state-scaled column density as a function of excitation temperature or energy above ground state) of ammonia emission in the Brick. Four temperature estimates(proportional to the inverse slope) are sketched and listed in the two boxes. Typical errors are 5%–10% in rotational and 10%–25% in kinetic temperature with increasing uncertainty for higher kinetic temperatures owing to theflattening of theTkin–Trotconversion (Table4; Morris et al.1973; Ott et al.2011; Gorski et al.2017).
intermediate mass black hole (Oka et al. 2016) because of its unusually high CO velocity dispersion.
The ammonia distribution as traced by the NH3 column density map closely follows 870μm dust emission obtained by ATLASGAL(Schuller et al.2009), as can be seen in Figure10 (bottom). The correlation is most pronounced for gas at higher column density than ~ ´1 1016cm−2. Lower column density
gas is mostly found in regions of weak dust emission, but the presence of dust does not imply the detection of ammonia at
´
Ntot 2 1015cm−2. Almost all (dense) ammonia gas is accompanied by dust emission. Notable exceptions from the overall good correlation are found in SgrB2 and SgrA*. SgrB2 (N) and SgrB2 (M) are detected in absorption in ammonia, whereas the dust emission peaks at these locations.
Figure 10.Top: total ammonia column density according to Equation(22). The dynamical range is ~ ´5 1015cm−2to ~6.8´1016cm−2. In warm clouds like SgrB2, this figure represents a lower limit as the unobserved states >J 6 can be populated in non-negligible fractions(Mills & Morris2013). Bottom: the same map of the total column density overlaid with 870μm dust emission from ATLASGAL (Schuller et al.2009) at contours of powers of two (1, 2, ..., 64) Jy beam−1. The ammonia distribution closely follows dust emission in regions of high ammonia column density( ´1 1016cm−2).
Figure 11. NH3(3, 3) line-of-sight velocity derived from fitting the hyperfine structure.
